MartinPascua/t1t0test

LORA basado en artista argentino

Descripción del Proyecto

Este proyecto es parte de la clase de Inteligencia Artificial Generativa para la carrera de Diseño. El objetivo es crear un modelo LORA (Low-Rank Adaptation) utilizando imágenes de un artista argentino como caso de estudio.

¿Qué es LORA?

LORA (Low-Rank Adaptation) es una técnica de fine-tuning para modelos de inteligencia artificial que permite adaptar modelos pre-entrenados a tareas específicas de manera eficiente. Esta técnica reduce significativamente el número de parámetros entrenables, lo que resulta en un proceso más rápido y con menos requerimientos de recursos computacionales.

Detalles del Modelo

• Plataforma utilizada: TOAST AI
• Número de imágenes: 8
• Tipo de imágenes: Obras de un artista argentino
• Propósito: Caso de estudio académico
• Uso comercial: No
• Palabra Clave: t1t0test

Proceso de Creación

El modelo LORA fue creado utilizando la plataforma gratuita TOAST AI. Se utilizaron 8 imágenes de un artista argentino para entrenar el modelo. Este proceso nos permitió adaptar un modelo de IA generativa pre-existente para que pudiera generar imágenes en el estilo del artista seleccionado.

Consideraciones Éticas

Es importante destacar que este proyecto se realizó únicamente con fines educativos y de investigación como parte de nuestros estudios en Inteligencia Artificial Generativa. No se busca ningún beneficio comercial y se respetan los derechos de autor del artista cuyas obras se utilizaron como base para el entrenamiento.

Limitaciones

• El modelo está basado en un conjunto limitado de imágenes (8), lo que puede resultar en una variedad limitada de outputs.
• Al ser un proyecto académico, no se ha realizado una evaluación exhaustiva del rendimiento o posibles sesgos del modelo.

Conclusión

Este proyecto demuestra la aplicación práctica de técnicas de IA generativa en el campo del diseño, permitiéndonos explorar cómo podemos adaptar modelos existentes para crear contenido visual inspirado en estilos artísticos específicos.

Prompt

t1t0test This cartoon shows two characters at a small table with a crystal ball, set against a green background. A yellow lamp hangs in the upper right corner. On the left sits a character in blue clothing on a small stool. On the right is a fortune teller character wearing a green dress with white polka dots and a blue headscarf, seated in a chair. Their hands hover near the crystal ball on the orange-clothed table. The speech bubbles now contain:First bubble (fortune teller):'Veo un gran viaje' (Translation:'I see a great journey') Second bubble (client):'¿En serio?' (Translation:'Really?') Third bubble (fortune teller):'Ah, no... un gran traje' (Translation:'Ah, no... a great suit')

Prompt

t1t0test The image shows a cartoon illustration with a green background, reminiscent of a soccer field. Two stylized male characters are sitting on a bench, likely on the sidelines of a soccer field. The character on the left is wearing a red and white striped jersey and shorts, while the character on the right is wearing a blue and white jersey and shorts, suggesting they support different teams. The characters have exaggerated features with large noses, similar to the original cartoon style. They are facing each other, appearing to be in the middle of a conversation. Between them, on the ground, is a soccer ball. Above them is a speech bubble containing text in Spanish that reads:"CREÍ QUE ERA TALENTO PERO FUE MARKETING, NOMÁS" The illustration maintains a simple and whimsical style with bold outlines and flat colors. In the background, there are minimal details suggesting a soccer stadium, such as the outline of goal posts and a scoreboard. This cartoon seems to be commenting on the nature of perceived talent in professional soccer, adding a humorous twist to a conversation between two friends who support rival teams.

Prompt

t1t0test This cartoon depicts a therapy or counseling session. The image has a beige background and shows two characters. In the foreground, a patient is lying on a gray couch or therapy bed. This character has a large nose and is wearing a purple shirt and blue pants. They are speaking, with a speech bubble above them that now says:"Acá mi perro no me deja dormir" (Which translates to:"Here, my dog doesn't let me sleep") In the background, seated in a chair, is a silhouette of what appears to be the therapist or counselor. They are holding what looks like a notepad and pen, presumably taking notes during the session. The overall style of the drawing is cartoonish and minimalist, with simple lines and shapes used to convey the scene. There's a signature or artist's mark in the bottom right corner that looks like "TJ" or "JC"

Prompt

t1t0test This cartoon shows a conversation between two characters against a light blue background. There's a lamp visible in the upper left corner and what appears to be a picture frame on the right wall. The taller character on the left has a large nose and is wearing a red shirt and blue pants. They are speaking to a smaller character, likely a child, who has blonde hair and is wearing a green outfit. The speech bubbles now contain:Taller character:"¿Quieres un helado de chocolate para tu cumple?" Smaller character:"No, uno de fresa con chispas"

Prompt

t1t0test This cartoon depicts two characters having a conversation in what appears to be a classroom setting. There's a green chalkboard visible on the left side of the image and a light fixture at the top. The character on the left is wearing a white coat, likely representing a teacher. The character on the right is wearing a blue suit, possibly representing a school administrator or principal. The speech bubbles now contain:Character in white coat:"Como maestro, me preocupa que los alumnos pasen mucho tiempo en sus celulares" (Translation:"As a teacher, I'm concerned that students spend too much time on their cell phones") Character in blue suit:"Tranquilo, use eso para enseñarles programación"

Trigger words

You should use t1t0test to trigger the image generation.

Download model and use it with ComfyUI, AUTOMATIC1111, SD.Next, Invoke AI, Forge, etc.

Weights for this model are available in Safetensors format.